Red Hat reveals major enhancements to Red Hat Enterprise Linux AI

Properly, that was quick. It was solely again in early September that Red Hat launched Red Hat Enterprise Linux AI (RHEL AI) 1.0. Now, Crimson Hat has adopted up by saying the overall availability of RHEL AI 1.2. This newest launch introduces a number of key enhancements aimed toward streamlining the event, testing, and deployment of enormous language fashions (LLMs).

Earlier than leaping into what’s new and improved, here is a bit about what RHEL AI brings. It is designed to streamline generative AI (gen AI) mannequin improvement, testing, and deployment. RHEL AI can be meant to make coaching LLMs reasonably priced. 

Additionally: Ubuntu 24.10 Oracular Oriole takes flight

This new platform combines IBM Analysis’s open-source-licensed Granite large language model (LLM) household, the LAB methodology-based InstructLab alignment tools, and a collaborative strategy to mannequin improvement by way of the open-source InstructLab AI venture.

RHEL AI additionally makes use of Retrieval-Augmented Generation (RAG) to allow LLMs to entry authorised exterior information saved in databases, paperwork, and different information sources. This strategy enhances RHEL AI’s skill to ship the precise reply relatively than a solution that sounds appropriate. 

Additionally: We’re a big step closer to defining open source AI – but not everyone is happy

This subsequent era of RHEL AI boasts expanded {hardware} assist. The brand new model now helps Lenovo ThinkSystem SR675 V3 servers, providing manufacturing facility preload choices for sooner and simpler deployment. In a expertise preview, RHEL AI 1.2 introduces assist for AMD Instinct Accelerators, together with the MI300x GPUs for coaching and inference and MI210 GPUs for inference duties.

RHEL AI 1.2 has additionally prolonged its attain to main cloud platforms — customers can now deploy RHEL AI on Azure and Google Cloud Platform (GCP), along with present assist for AWS and IBM Cloud.

Additionally: Linus Torvalds talks about AI, Rust adoption, and the Linux kernel

The software program has additionally seen important enhancements. The brand new “Periodic Checkpointing” characteristic saves lengthy coaching runs at common intervals throughout fine-tuning. Customers can resume coaching from the final saved checkpoint as an alternative of beginning over, saving priceless time and computational assets.

RHEL AI 1.2 comes with the PyTorch Fully Sharded Data Parallel (FSDP). Launched as a expertise preview, FSDP dramatically reduces coaching instances for multi-phase coaching of fashions with artificial information. The expertise shards a mannequin’s parameters, gradients, and optimizer states throughout parallel employees (for instance, GPUs), chopping coaching instances by appreciable quantities.

Additionally: How open source is steering AI down the high road

RHEL AI continues Crimson Hat’s transfer to make LLM coaching extra accessible to programmers and material consultants, not simply information scientists. 

As Joe Fernandes, Crimson Hat’s Basis Mannequin Platform vp, mentioned: “RHEL AI supplies the power for area consultants, not simply information scientists, to contribute to a built-for-purpose gen AI mannequin throughout the hybrid cloud whereas additionally enabling IT organizations to scale these fashions for manufacturing by way of Crimson Hat OpenShift AI.” 

In different phrases, with RHEL AI, making AI helpful in your specific wants is changing into even simpler. 

Lastly, with the discharge of RHEL AI 1.2, Crimson Hat can be deprecating assist for model 1.1, giving customers 30 days to improve to the newest launch. This fast iteration underscores Crimson Hat’s aggressive push into the enterprise AI market. For higher or for worse, AI development is accelerating at an ever-increasing tempo.